Valar
Valar is the most efficient, lowest-cost inference layer for agentic workloads across GPUs, accelerators, and clouds.
Valar sells inference on Liquid Inference. A buyer names one of its models as valar/<model>, or names the model and pays the lowest offer that meets the buyer's rules.
Facts
- Handle
- valar
- Website
- https://valarhq.ai/
- On the exchange since
- 2026-10-20
- Models
- 2
- Offers
- 2
- Regions
- us-east
- Precisions (declared)
- fp8
- Lowest price on
- 0 of 2 models
Reliability, last 3 days
- Probes answered
- 97.32%
- Hours with an answer
- 100.0%
- Median response time
- 1.3s
Checked by the exchange
1 of 2 offers match the fingerprint of the model. 0 have a finished evaluation.
The exchange sends fixed prompts to each offer and compares the answers with a baseline for the model. It also runs benchmarks against the offers and keeps the results.
Data handling
- Zero retention, declared by the provider
- No training, declared by the provider
- SOC 2 Type II, declared by the provider
- GDPR DPA, declared by the provider
- PCI DSS, declared by the provider
Models
- DeepSeek V4.1 Flash 1017: declared precision fp8, us-east, $0.048 input and $0.178 output per million tokens, $0.0048 cache read per million tokens, 1M tokens of context
- GLM 5.3: declared precision fp8, us-east, $0.800 input and $2.500 output per million tokens, $0.100 cache read per million tokens, 1M tokens of context